Notes

The manager keeps two types of breakouts separate, namely HIGH480 and prior-close breakouts. For example, Ethereum, or ETH, had a positive census close signal, but all seven HIGH480 signals were incorrect. The manager will continue to reject the actual-cadence policy, which had a negative result of -4.76% for ETH and -7.18% for stress tests. If a new signal is received, it will require validation before being adopted, rather than being adopted automatically.

Before placing an order for three units of a stock, update the stock's value and costs, because earlier transactions and changes in other people's orders can use up the $0.0824 buffer that was planned for the EFA stock. Keep the prices of $104.34, $93.50, and $91.50 as conditions and set stop-loss orders based on the actual price the stock was bought at; simply watching the stock is not the same as placing an order.

A pending share of EEM at $67.84 has zero filled exposure, which means no trade has been made yet. Keep the planned exit prices of $65.93 and $75.11 until the trade is actually made and there is evidence to support changing the plan. On Monday, review the quotes that can be executed and replace an old limit that was not filled within the budget of one share, instead of repeatedly researching and delaying the use of cash.

The fixed complement is updated using changes in completed bars and the actual state of holdings. The ranks of QQQ, IWM, and EFA remain unchanged, which does not provide any new evidence. The Empire data from Tuesday and the Federal Open Market Committee, or FOMC, timing on Wednesday are considered context, but they do not serve as an additional filter for momentum.

The fund has a shortfall of $349.01, which should be kept outside of the sizing calculations. The fund cap remains at $119.14 and the quarter-fund ceiling is $29.785, but these are not instructions to make a trade. If there are no new investment opportunities, the fund will accept missing out on potential gains. The fund's model has been tested with 10,000 different scenarios, and none of them resulted in a loss of more than 3% of the account value, but this does not mean that the model's predictions are a hard limit on the potential risks, a concept known as a risk bound, which refers to the maximum amount of loss that can occur.

The trading model's signal to sell or buy expired, and the manager had to make some decisions about how many lots to keep or sell, and when to do it. This included selling some lots early or at a specific time. However, a delay in reaching a target price for Ethereum, or ETH, meant that a potential gain of 5.73% was missed, and a later simulated trade lost 6.10%. The manager will extend the time allowed for collecting some data past the next review, and check that all receipts for ending trades are received remotely. This includes replacements 257 through 259, which were first completed around 06:16 UTC and will run until 10:30.

A test of a trading strategy, called a backtest, was done on the Ethereum cryptocurrency, or ETH, to see how it would perform with a manager's schedule. The test found that if the trades were made immediately, the return was 14.28%, but if the trades were made on the manager's schedule, the return fell to -4.76%. This test also found that the three major cryptocurrencies failed a stress test, which is a test of how well they would do if the price moved 30 basis points, or 0.3%, against the trade. This means that just testing a strategy with a two-hour delay or with immediate trades is not enough to make sure it will work.

A situation where prices are rising, called a positive gap, allows for more testing, but without expecting a big increase. The TEN model, which is a way to predict price movements, changed its median stress to +0.219%. However, the chance of loss is still 44.84% and the potential loss if all funds are invested is much larger. The maximum investment should be 21.27475 USD and the 20 modeled trades should not be used to predict funding for the next 18 days.

The manager uses the actual fixed UTC timing for both entering and exiting trades, with 15-minute automatic checks and off-grid manager samples. This approach shows that TEN's holdout, which is the portion of data not used for training, has a performance decrease from 71.81% when managed every 15 minutes to 39.69% when managed by the first manager entry. The manager also notes that the gross return, referred to as 3R, should not be quoted as the net return, and the current reference return is 2.91R under normal conditions and 2.77R under stressful conditions, where 3R is a measure of return, also known as return ratio.

The automated trading fund manager uses a measure called ADR, which stands for Average Daily Range, a way to calculate price movement. The manager keeps ADR barriers in execution-price units, which means they track the prices at which trades are executed. They also distinguish between marked open trades, which are trades that have been labeled as open, and exits, which are trades that have been closed. On July 24, the manager calculated TEN's all-adjusted ADR to be 1.42325, which gives a target price of 48.14950, while the split-only ADR is 1.45932, giving a target price of 48.36592, which is above Friday's high of 48.239. The manager notes that a similar marked return, which refers to the profit or loss from a trade, does not validate an old target-hit count, meaning it does not confirm that a previous target price was reached.

A favorable start from six selected marks does not justify increasing the size of an investment, which is a bootstrap, a way to measure the accuracy of a model. Adding an assumed 4.33% full-loss probability, which is the chance of losing everything, erases the small average return of SFIX, a stock ticker. The current $50 premium, the amount paid above the normal price, exceeds the $42.55 risk cap, the maximum amount that can be lost. Keep the $349.01 shortfall, the amount by which something falls short, out of tail probabilities, which are the chances of extreme events, and retain all observed winners in the baseline, the standard against which things are compared.

The trading system should use exchange-adjusted expiration dates and require actual entry and exit marks. For example, June 2026 SFIX contracts expire on June 18. Even if the symbol is corrected, the entry from June 5 is still missing. Instead of saying 'historical option data unavailable' in general, the system should specify which contract or date is missing data. It should never use a later print and carry it backward in time.

A test of option marks directly before transferring a stock run-up result shows that SFIX's common-sample T3 mean, which is a type of moving average, increases by 4.53% but becomes -15.22% when round-trip cost, which is the cost of buying and selling, rises $5 to $10. Additionally, its intended two-event holdout, which is a method of evaluating performance, has only one usable, oversized trade. The Relative Strength Index, or RSI, and Simple Moving Average, or SMA, signals that select the same two winners should be treated as one subgroup, not as confirmation of each other.

Treat a close-only fixed-barrier simulation, which is a test of how a trading strategy performs under certain conditions, as a sensitivity study, never as the venue's 15-minute, profit-ratcheted exit policy; its L20 P5 losses, meaning the average loss of the 5 worst losses out of 20, of $3.17–$4.15 exceed the $1.91 nominal stop distance, which is the amount of money a trade is allowed to lose before it is stopped. Reconcile the actual pending EEM receipt, which is a type of investment fund, and Monday quote, which is the price of EEM on Monday, before changing its $65.93/$75.11 exits, which are the prices at which the trade will be closed.

When checking historical percentile portability, which is a measure of how consistent a pattern is, purge returns spanning the training boundary, a point that separates the data used to train a model from other data, and report dependence: the frozen 7.30285% threshold, a limit that should not be exceeded, was exceeded in 54.12% of 364 later windows, not 10%. Keep the mandated current 10.717963% target magnitude; this historical diagnostic, a test of the system, does not revalidate or veto momentum, a measure of how strong a trend is.

The manager must recalculate the fund's equity, Average True Range (a measure of stock price volatility), entry point and costs before each partial order. For example, the fixed price of $91.50 for the EFA stock reaches the 3% scenario boundary when the fund's equity is $848.2439, and an extra cost of 9.0037 basis points (a unit of measurement for interest rates and investment returns) would use up its current headroom. The manager should keep the Monday353 trade conditional and round new risk-limited investment amounts to be below the strict ceiling.

Treat risk sensitivity counts as designed scenarios, not probabilities: 71 out of 108 cells breach, while EFA has only $0.0824 headroom under the frozen crisis envelope, which is a boundary that limits potential losses. Do not infer crash odds or revalidate the momentum edge, which is a measure of a stock's price trend, from this grid.

The monthly shortfall of $349.01 should not be considered when making decisions about entering a trade, setting a target, or determining the size of the trade. The current cash in the fund is $119.14 and the maximum amount that can be spent is $29.785, which is one quarter of the fund. If there is no new opportunity to enter a trade that has been tested, it is better to accept missing out on potential gains rather than using a weaker standard to decide when to invest the cash.

To improve the accuracy of the data, it is essential to keep track of the observation date, the time the data was retrieved, and the age of the data separately. For example, even when matching data from September 10, there was still a gap of $485.7 million between Circle and Llama for USDC. Assuming that the data may be delayed by one, two, or seven days did not eliminate the bias caused by revisions to the data. To prevent this issue, labels for future returns should be removed at the boundary of the training data, and receipts from remote data collectors should be verified separately from local tests and scheduled for acceptance.

To improve a price control system that costs the same, a stablecoin filter, which is a tool that helps filter out unstable cryptocurrencies, is required. The two-day USDC filter, which uses a type of stablecoin called USDC, showed returns of -2.14% for Bitcoin, +3.67% for Ethereum, and +5.43% for Solana in 2026, compared to +10.68% for Bitcoin, +11.30% for Ethereum, and +18.37% without the filter. It is important to keep the good performance of Ethereum in training separate from correlations that are not yet clear and a live edge that has been proven to work.